Dodd, V. from The Guardian reported that 2 Randox Testing Services (RTS) employees are under criminal investigation for suspected manipulation of forensic tests. 6,000 forensic samples are to be re-examined as part of the investigation, however 10 per cent of samples are no longer available. These 10 per cent of samples that cannot be relied on, therefore cases involving these samples might get overturned.
It is, however, known that the samples re-tested so far are showing the same results as before, meaning that the manipulation might not have been successful or might not have affected the results. It is simply too early to say. We need more re-tested samples to compare with the previous results to establish a probability of error. Samples where an error occurred can be cross-examined in order to find similarities. Common indicators of error can be cross-referenced with cases without samples and a probability of the original test being unreliable can be established based on specific conditions, such as biological traits. Samples where the original test and re-test show different results and where no common indicators are found should be included when calculating probability of error for all cases without a sample to re-examine.
Of course, such approach would not provide the reliability we usually expect from a forensic test, but it would still indicate a probability of having a reliable test, which may still be higher than the reliability of eyewitness evidence, for example.
Forensic evidence is the safest form of evidence when used correctly. But that does not mean we do not use other types evidence, we just use caution and evaluate the likelihood of it being true. Therefore I think we should do the same here – wait for a large sample size and then apply statistical analysis to evaluate the reliability of previous tests.
You can read the article about the investigation on The Guardian website here: https://www.theguardian.com/uk-news/2017/may/09/6000-forensic-samples-re-examined-in-inquiry-into-manchester-lab